Hey, > Perhaps I mis-wrote. 90% of bcachefs could be ripped out and run in > userspace without FUSE. 10% has of bcachefs has linuxisms. Kent > really pushes for making it possible to run code in both kernel space > and userspace. He gave an example of some code that deals with a > creating datastucture in Linux...that should be possible to export > as a userspace library really easily. We spoke with him about porting > it natively to the Hurd (1).
Thats still 10%, and considering the amount of features bcachefs has (as intended to be a openzfs/btrfs competitor) I am sure 10% is a huge amount of code. > A native port to OpenBSD (maintained in base) is not possible. > Someone could fork OpenBSD (call it LibreBSD) and maintain bcachefs > in base... Forking like this is exactly the issue Linux has, where you have 101 distros all using the exact same codebases, but disagreement on how they should be run or what packaging format they should use. BSDs are a family of operating systems, and each one of them has their own pros and cons, it is not like Linux where you die on a hill defending your favourite distro, but more applying the right BSD to the right problem. The best chance of bcachefs on BSD would be FreeBSD, which are happy to support big filesystems such as openzfs, and they have more developers, and a lot more financial backing than OpenBSD does (to my knowledge). You have to also bare in mind that OpenBSD believes in proactive security and auditing the code often. Adding complexity to it then makes it much harder to audit, and much more likely for a security flaw to go unnoticed. > But I thought FUSE was meant to be portable accross filesystems. > Shouldn't it be possible to run any FUSE filesystem on any os that > supports FUSE? I am not a kernel developer, I can not say. But what I can say is bcachefs likely has the same issue openzfs has which is it has a huge complex codebase which makes it difficult to audit (and port), and there is likely not the manpower for OpenBSD to port bcachefs even if it wanted it. I am not too sure the drama surrounding Linux and bcachefs, I do not pay attention to Linux news very much anymore because I deem Linux to a large extent, a lost cause. However, the fact that bcachefs no longer has in-kernel support, and is now a DKMS module means that not even Linux is likely to have reliable support for bcachefs. It doesn't seem like anything has changed, Linux has btrfs, FreeBSD/NetBSD has openzfs, if OpenBSD wanted to have a complex filesystem they would have ported openzfs. (correct me if I am wrong) I do not see the purpose of the thread, unless you are willing to port the bcachefs filesystem over to OpenBSD. Are you asking for someone else to do the hard work or to help? Nothing is stopping you doing it yourself if your theory of it being possible is true, however I heavily doubt anyone on the mailing list shares your views, bcachefs is GPLv2-only which violates the licencing policy, and it doesn't fit OpenBSD security model, and thats before even discussing the complexity or the amount of labour is required to port it. Take care, -- Polarian Jabber/XMPP: [email protected]
